<?xml version="1.0" standalone="yes"?>
<Paper uid="M93-1024">
  <Title>Tagged Sentences Tagger Lisp-readable sentences CompletedChart Template Generator, Target Templates Input Text System Knowledge Bases</Title>
  <Section position="7" start_page="301" end_page="302" type="concl">
    <SectionTitle>
CONCLUSION
</SectionTitle>
    <Paragraph position="0"> Although LINK's performance on the English Microelectronics testset was less then stellar, it is difficult to draw conclusions about the use of our approach on this domain . The primary reason for degradation of performance as compared to MUC-4 was a lack of resources needed t o develop a proper knowledge base for the domain . Lack of information in the lexicon, grammar , and system's domain knowledge resulted in poor analysis of the majority of articles .</Paragraph>
    <Paragraph position="1"> The system's relatively good performance in precision indicates that our full-analysis approach is likely to yield reliable results when information is extracted . However, the Microelectronics domain may be unusual in the frequency of technical terms which are not commonly use d in general English. Because of this property of the domain, it appears that techniques relying o n less complete analysis of the text may be appropriate also . Our own experience indicated that the utilization of even partial parses of only a few words (or even a single word) improved ou r system's recall without damaging precision . We believe that in a domain with a less specialize d vocabulary, techniques relying on specific keywords would be more likely to degrade precisio n performance.</Paragraph>
  </Section>
class="xml-element"></Paper>